Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
Although not as safely as SMB1, I think it's very likely that [3989] SNES Super Mario World "game end glitch" by Doomsday31415, BrunoVisnadi, Masterjun in 00:41.68 cannot be pushed further.
However, I also used to believe it for the previous 2 publications of this category (and this hypothesis was actually my motivation to try to improve it in first place), so I see some decent chances someone eventually proves me wrong
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
p4wn3r wrote:
The number of deaths, at least those evaluated here, includes anyone who test positive for COVID. I am fairly certain of that, because I know someone who died from a stroke, but because he had tested positive, he was counted as a COVID death, and there were sever restrictions on the mourning procedures.
I can't tell what happened to the person you knew, though if they tested positive for covid, keep in mind it's a strong stroke inducer.
Nevertheless, there is a difference in Brazil between the registration of a covid death at the obituary and how these deathes are actually counted. Most deaths that envolve someone who tested positive or who were suspicious of having covid are registred as covid deaths simply to ensure protective measures when moving the body and at the funeral conduction. However, these deaths aren't counted to the country numbers if there is no further confirmation.
p4wn3r wrote:
About this pandemic? I can say the hospitals are not nearly as chaotic as they were during some Dengue outbreaks we had in the past. I still cannot understand how this disease can need strict lockdown actions, while at the same time be very difficult to detect in your daily routine (except, of course, through the media, which has been airing COVID numbers nonstop for the last five months).
As opposed to Dengue, this disease spreads along crowds through human contact. Also, it's been killing at a 300x higher rate than Dengue was in it's worst years.
If our hospitals didn't get as bad as the ones we say in Italy, it might have been because authorities decreed lockdown measrures early enough.
Still, even if the deaths were suficientelly spread out (temporally and geographically) so that no particular hospital would find itself in a completely chaotic situation, the lives saved pay off enough for the unpleasant measures like social distancing.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
Lil_Gecko wrote:
Hello everybody.
Can somebody help me figure out how to do the spawn Reznor glitch ?
I have Yoshi in slot 6
P-Switch in slot 8
2 horizontal fishes in slot 9 and 7 in water.
I do the double-tongue glitch, once on the P-switch, and catch-cancel the Fish in slot 7 with the second.
So I think that now the fish is stunned right ?
But when I spit it out, nothing happens, no Reznor, it's just the fish flopping around.
I've seen a video performing it in this exact level so I know it's possible. What am I missing here ?
Thanks.
You probably realized it since you later said it worked, but you also don't want to use a fish in slot 7 since this is the slot the Reznor must take in order to it to function properly.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
I think it's up to you guys. You might cancel it or set it as delayed, not sure what's more appropriate. I can't guarantee how early I'll be able to come up with the improvement
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
Amaraticando wrote:
" the GBA version does not have a ceiling, so we can get as high as we can in this area."
Then Mario should get as high as possible in order to minimize input length, right?
The reason for which I didn't submit this movie I made years ago is precisely due to the fact it would be ridiculously hard to end input optimally, for various reasons. It's way worse than it looks, it'd take weeks and tons of effort, so I simply got demotivated.
Colin synced my movie to Bizhawk some time ago, which is something I appreciated, and he insisted to me quite a lot to let him submit it. At first I didn't want him to because I really don't like to send material I feel to be suboptimal, but I eventually agreed because I realized I'd never get around to touch that movie again and it's probably better to submit something than to submit nothing at all. After seeing the Bowser Kill, though, I don't know how I feel about this anymore :/
I thought Colin would aim for fastest Bowser Kill instead of fastest input end (I think he said that at some point but I might be wrong), which I would be OK with, although it's not my favorite style. Watching now, it seems he didn't go for any of these goals at all :/
It obviously makes no sense to hurt Mario if you are aiming for fastest kill (no point in letting the screen to freeze for a while), and if you are aiming for shortest input Mario should have started flying 30 seconds sooner, carrying a Mechakoopa.
Since most of my real life activities got cancelled due to covid, I'll have much more free time than usual for a few months. Thus, I'll be able to do some TASing during this time, and although I mostly intend to work on 96 exit, I might touch this movie again and give it a proper end of input. In a couple weeks, hopefully I will get it done.
I'm sorry if this sounds harsh to Colin, but as I said, I really don't like the position of having suboptimal material submitted. It's related to how I view the 'philosophy' of TASing and what I think the point of this should be, but we can discuss this (and we probably did several times) at a more appropriate place.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
Nach wrote:
BrunoVisnadi: You said you had an SRAM bug to look into? Any progress?
Ops, I forgot I left that message. If you were waiting for me to check it to proceed judging, I apologize.
I talked to a couple people and they said it's probably a parser bug, so I didn't bother diving into the details. If it's crucial for the acceptance of the submission, I might look into it and/or ask for more help, but I will only have time for so in a couple weeks at least.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
warmCabin wrote:
Idk what you mean. Pretty sure I saw some rune or symbol like that once, in like Stone Henge or something.
Edit: found it
It's what Colin said in the post before this one. Interesting observation with the similarity with the runes though.
HappyLee wrote:
Note: It seems that the new TAS saves 3 frames in the Dark Room, but somehow loses 1 frame in the final Bowser fight.
I didn't clarify this in the submission text. In the previous run, we slowed down a bit in the dark room to reduce lag, so we entered Bowser's door 1 frame late, which was the good frame parity. This time the best way I found to reduce lag involved losing less subpixels. I entered the door, thus, 1 frame before, and saved 2 lags, 3 frames in total. Only the lags mater, though, since Bowser started with bad frame parity and I immediately lost that frame back.
Thanks for the comparison video, by the way!
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
warmCabin wrote:
Why not do YI1 first? Wouldn't that save 8 units of backtracking?
It would be impossible to stun glitch in YI4, because it's necessary to take damage and the orb would fall from the item box.
warmCabin wrote:
In YI3, how is the red shell contributing to the dupes?
When a turn block is hit by the first time in a level, it starts spinning and a block sprite, which takes slot 3, is set to it. If you hit more turn blocks before the first stops spinning, they'll start spinning as well and their block sprites will take slot 2, then 1, then 0. If none of the 4 slots is slots available, however, and you hit a new turn block, the block that is on slot 3 will stop spinning, and the block you just hit will start spinning, taking slot 3. If you keep doing this, then slot 2 will be taken, then 1, then 0, then 3 again. That is, it's impossible to 5 blocks to spin at the same time.
What was done in previous runs was to, after duplicating each block, double hitting (hitting with Mario and with a item at the same time) 2 blocks quickly, so that 4 sprite blocks are spawned and the block you duplicated stops spinning and you can duplicate it again. Instead of that, I used this method with the red shell, so that while Mario was hitting blocks around the duplications, the shell was contributing hitting blocks much faster than Mario could do by himself.
warmCabin wrote:
Are those some kind of runes?
Well, that was a reference to a meme that came a couple years too late.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
Thank you all for the positive feedback!
The8bitbeast wrote:
Interesting about the cloud glitch too. How much time do you think cloud would save if it were used?
It skips most of the Bowser fight, but you can't orb the Castle or stun glitch in YI4. Perhaps ~40 seconds?
Challenger wrote:
What impressive how the amount of frames saved is the same as the previous TAS. Maybe because of framerules?
It's just a nice coincidence. The star warp frame rule does force any improvement of real frames to be a multiple of 8, but there are tons of lag improvements and trade-offs in this run as well, which aren't affected by frame rules.
Challenger wrote:
Also, how the progress of 96-exit is going?
About half way through, it's been a while I don't work on it. Most likely scenario is I will give up of some very early improvements and of deep score/coin/frame rule planning in order to get it finished some day.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
p4wn3r wrote:
Now, for general N, it's easy to see if you solve for N prime, you solved for general N.
I don't find it that obvious. It's easy to see that by solving for all primes you solve for powers of primes, but how about products of different primes? Not that this maters if for all N there is a Pascal Triangle row with the property you conjectured. Looking at small values of m, though, I couldn't find a row with all elements multiples of 4.
Also, nice work on the summation problem I posted. I hadn't considered picking b so that b/(2pi) is rational. The most challenging scenario is when it isn't, but after seeing your input I don't know if I'll adventure myself into trying to prove anything here.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
Does the following sum always converge, for all non-zero values of a and b?
What if n, in the denominator, is raised to c, 0<c<1? Can this sum diverge for small enough values of c?
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
My point was that the top vertex being above the centroid requires t to not be too small. If a, b and c are just slightly bigger than l/2, I believe the construction is impossible
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
I believe if we take a = b = c = l/2 + t, for a very small t, it's not hard to see it's impossible to construct the pyramid, unless there is something I understood wrong. The top vertex would need to be in the intersection of 3 regions that don't intersect at all.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
As I said on Discord, I found a better strat for the item swap as I was testing this. I believe it must save more than 5 frames. However, I couldn't get the credits with Masterjun's inputs.
I'm on phone right now so I can't give the movie file. It's in #tas in SMW Discord.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
Very nice find, I'm surprised we missed that.
It should be possible to fix this by changing the parity of the green shell's slot. I'll try to make you a movie of it.
EDIT: Nevermind, the shell must be in slot 9 in order to the eat swap to work. I'll still try to figure out a solution.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
Nice puzzle. I don't have the time to put much thought into it right now, but I found it to be useful to think of each hero as a point in the plane. If you arrange them in a square, it's easy to see that N = (X+1)^2 is an upper bound for the amount of heroes the Resistance must send, though it's sub-optimal even for X = 1.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
I disagree with some of the predominant views on this thread. Assuming the re-record count displayed on the submission is correct, I believe it is a good measure of how much effort was put into creating the run and it says something about its technical quality.
Of course, this heavily depends on the game, usage of TAStudio, how the emulator counts re-records, etc. I'm talking about the most common contexts. When I see a SMW submission with a low re-record account I can tell it's not optimal before watching it, and so far I have never been wrong about it.
We are talking, roughly, about the amount of times the author edited their submission, and the fact is it's simply impossible to create an optimal movie with a tiny number of editions. Similarly, if the number of editions is huge, it's safe to assume the author was very careful about saving frames and their run is optimal to some extent. I don't think the submission text can make up for the information provided by the re-record count, as I've seen long submissions texts of runs with few re-records that weren't really optimal. Additionally, some excellently optimized TASes can have small submission texts because the author doesn't want to get into too technical details.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
Most of this was posted as a message on Discord to the author of the run, regarding possible improvements, but I should post it here too:
Firstly, congratulations for your submission. It's significantly better than your previous ones, especially in the first levels. It's a sign you are improving, but there are still some mistakes you should pay attention to.
At the beginning of 2nd level you walked left for many frames, which loses time. The -1 trick consists of pressing left for a single frame. I noticed you forgot to do it in some levels, like level 3.
At the 4th level, I feel it's probably possible to skip a cycle if you use the enemies in your favor to jump over lava. It should save time at the end too. The pipe entry wasn't good, you lost a lot of speed on the ground for many frames.
At the first room of the castle you lost a lot of time on the grid. Mario moves incredibly slowly on grids, you shouldn't touch them for more than 1 frame. It should be tested if it's possible to skip the grid entirely by spin jumping the podoboo with P-Speed. You also lost P-speed at the end of the room, which is probably avoidable. Even if it isn't, it's way better to 6/5 instead.
At the second room, you could've made an enemy to move left and spin jump on it instead of waiting for the platform. There was also a lot of delay to enter the boss door.
At the boss, visually it feels like it's clearly possible to hit him sooner, but I might be mistaken. Anyway, hitting the ball once definitely loses time.
About minor stuff, getting bonus starts is usually a bad idea, because they cause the game to lag more often. But there is much more into ending levels optimally, like score/coin planning, manipulating Mario's subspeed, delaying goal tape end when the frame rule is bad in order to save lag, etc. I believe the most significant minor improvement you could have applied is to partially clip horizontal pipes as you enter them (like in an walljump), which shortens the animation a good amount. Corner boosting can also save some frames here and there.
I believe you aren't willing to do yet another version of this run, but even if you are, this hack doesn't seem to be good or interesting enough for acceptance on TASVideos. Anyway, these points are mentioned are relevant if you want to TAS a different SMW hack in the future. Good luck with your next project.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 12/28/2013
Posts: 396
Location: Rio de Janeiro, Brasil
I don't know a lot about SMB, but I think you don't need to slow down that much to get a good subpixel. I don't see why you would need to lose more than 1 frame slowing down.
EDIT: Nevermind, I rewatched it and understood better what happened. I'm confused about the last clip, though. Couldn't it be done quicker?
What if n, in the denominator, is raised to c, 0<c<1? Can this sum diverge for small enough values of c?